00001 /*
00002  * cpu_accel.c
00003  * Copyright (C) 2000-2002 Michel Lespinasse <walken@zoy.org>
00004  * Copyright (C) 1999-2000 Aaron Holtzman <aholtzma@ess.engr.uvic.ca>
00005  *
00006  * This file is part of mpeg2dec, a free MPEG-2 video stream decoder.
00007  * See http://libmpeg2.sourceforge.net/ for updates.
00008  *
00009  * mpeg2dec is free software; you can redistribute it and/or modify
00010  * it under the terms of the GNU General Public License as published by
00011  * the Free Software Foundation; either version 2 of the License, or
00012  * (at your option) any later version.
00013  *
00014  * mpeg2dec is distributed in the hope that it will be useful,
00015  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00016  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
00017  * GNU General Public License for more details.
00018  *
00019  * You should have received a copy of the GNU General Public License
00020  * along with this program; if not, write to the Free Software
00021  * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
00022  */
00023 
00024 #include "config.h"
00025 
00026 #include <inttypes.h>
00027 
00028 #include "mpeg2.h"
00029 
00030 #ifdef ACCEL_DETECT
00031 #ifdef ARCH_X86
00032 static inline uint32_t arch_accel (void)
00033 {
00034     uint32_t eax, ebx, ecx, edx;
00035     int AMD;
00036     uint32_t caps;
00037 
00038 #ifndef PIC
00039 #define cpuid(op,eax,ebx,ecx,edx)       \
00040     __asm__ ("cpuid"                    \
00041              : "=a" (eax),              \
00042                "=b" (ebx),              \
00043                "=c" (ecx),              \
00044                "=d" (edx)               \
00045              : "a" (op)                 \
00046              : "cc")
00047 #else   /* PIC version : save ebx */
00048 #define cpuid(op,eax,ebx,ecx,edx)       \
00049     __asm__ ("push %%ebx\n\t"           \
00050              "cpuid\n\t"                \
00051              "movl %%ebx,%1\n\t"        \
00052              "pop %%ebx"                \
00053              : "=a" (eax),              \
00054                "=r" (ebx),              \
00055                "=c" (ecx),              \
00056                "=d" (edx)               \
00057              : "a" (op)                 \
00058              : "cc")
00059 #endif
00060 
00061     __asm__ ("pushf\n\t"
00062              "pushf\n\t"
00063              "pop %0\n\t"
00064              "movl %0,%1\n\t"
00065              "xorl $0x200000,%0\n\t"
00066              "push %0\n\t"
00067              "popf\n\t"
00068              "pushf\n\t"
00069              "pop %0\n\t"
00070              "popf"
00071              : "=r" (eax),
00072                "=r" (ebx)
00073              :
00074              : "cc");
00075 
00076     if (eax == ebx)             /* no cpuid */
00077         return 0;
00078 
00079     cpuid (0x00000000, eax, ebx, ecx, edx);
00080     if (!eax)                   /* vendor string only */
00081         return 0;
00082 
00083     AMD = (ebx == 0x68747541) && (ecx == 0x444d4163) && (edx == 0x69746e65);
00084 
00085     cpuid (0x00000001, eax, ebx, ecx, edx);
00086     if (! (edx & 0x00800000))   /* no MMX */
00087         return 0;
00088 
00089     caps = MPEG2_ACCEL_X86_MMX;
00090     if (edx & 0x02000000)       /* SSE - identical to AMD MMX extensions */
00091         caps = MPEG2_ACCEL_X86_MMX | MPEG2_ACCEL_X86_MMXEXT;
00092 
00093     cpuid (0x80000000, eax, ebx, ecx, edx);
00094     if (eax < 0x80000001)       /* no extended capabilities */
00095         return caps;
00096 
00097     cpuid (0x80000001, eax, ebx, ecx, edx);
00098 
00099     if (edx & 0x80000000)
00100         caps |= MPEG2_ACCEL_X86_3DNOW;
00101 
00102     if (AMD && (edx & 0x00400000))      /* AMD MMX extensions */
00103         caps |= MPEG2_ACCEL_X86_MMXEXT;
00104 
00105     return caps;
00106 }
00107 #endif /* ARCH_X86 */
00108 
00109 #ifdef ARCH_PPC
00110 #include <signal.h>
00111 #include <setjmp.h>
00112 
00113 static sigjmp_buf jmpbuf;
00114 static volatile sig_atomic_t canjump = 0;
00115 
00116 static RETSIGTYPE sigill_handler (int sig)
00117 {
00118     if (!canjump) {
00119         signal (sig, SIG_DFL);
00120         raise (sig);
00121     }
00122 
00123     canjump = 0;
00124     siglongjmp (jmpbuf, 1);
00125 }
00126 
00127 static inline uint32_t arch_accel (void)
00128 {
00129     signal (SIGILL, sigill_handler);
00130     if (sigsetjmp (jmpbuf, 1)) {
00131         signal (SIGILL, SIG_DFL);
00132         return 0;
00133     }
00134 
00135     canjump = 1;
00136 
00137     asm volatile ("mtspr 256, %0\n\t"
00138                   "vand %%v0, %%v0, %%v0"
00139                   :
00140                   : "r" (-1));
00141 
00142     signal (SIGILL, SIG_DFL);
00143     return MPEG2_ACCEL_PPC_ALTIVEC;
00144 }
00145 #endif /* ARCH_PPC */
00146 
00147 #ifdef ARCH_ALPHA
00148 static inline uint32_t arch_accel (void)
00149 {
00150     uint64_t no_mvi;
00151 
00152     asm volatile ("amask %1, %0"
00153                   : "=r" (no_mvi)
00154                   : "rI" (256));        /* AMASK_MVI */
00155     return no_mvi ? MPEG2_ACCEL_ALPHA : (MPEG2_ACCEL_ALPHA |
00156                                          MPEG2_ACCEL_ALPHA_MVI);
00157 }
00158 #endif /* ARCH_ALPHA */
00159 #endif
00160 
00161 uint32_t mpeg2_detect_accel (void)
00162 {
00163     uint32_t accel;
00164 
00165     accel = 0;
00166 #ifdef ACCEL_DETECT
00167 #ifdef LIBMPEG2_MLIB
00168     accel = MPEG2_ACCEL_MLIB;
00169 #endif
00170 #if defined (ARCH_X86) || defined (ARCH_PPC) || defined (ARCH_ALPHA)
00171     accel |= arch_accel ();
00172 #endif
00173 #endif
00174     return accel;
00175 }